Report: Intel Could Spin Out Foundry Business or Cancel Some Expansion Plans to Control Losses

[AleksandarK]

According to a recent report from Bloomberg, Intel is in talks with investment banks about a possible spin-out of its foundry business, as well as scraping some existing expansion plans to cut losses. As the report highlights, sources close to Intel noted that the company is exploring various ways to deal with the recent Q2 2024 earnings report. While Intel's revenues are in decline, they are still high. However, the profitability of running its business has declined so much that the company is now operating on a net loss, with an astonishing $1.61 billion in the red. CEO Pat Gelsinger is now exploring various ways to control these losses and make the 56-year-old giant profitable again. Goldman Sachs and Morgan Stanley are reportedly advising Intel about its future moves regarding the foundry business and overall operations.

The Intel Foundry unit represents the biggest consumer of the company's funds, as the expansion plans across the US and Europe are costing Intel billions of US Dollars. Even though the company receives various state subsidies to build semiconductor manufacturing facilities, it still has to put much of its capital to work. Given that the company is running tight on funds, some of these expansion plans that are not business-critical may get scraped. Additionally, running the foundry business is also turning out to be rather costly, with Q2 2024 recording a negative 65.5% operating margin. Separating Intel Product and Intel Foundry may be an option, or even selling the foundry business as a whole is on the table. Whatever happens next is yet to be cleared up. During the Deutsche Bank Technology Conference on Thursday, Pat Gelsinger also noted that "It's been a difficult few weeks" for Intel, with many employees getting laid off to try to establish new cost-saving measures.



View at TechPowerUp Main Site | Source

 

[JWNoctis]

Large room. I hear an echo.

Bad news. Best of luck to them, whatever comes.

 

[john_]

If Intel doesn't come out of this mess intact in a year or two, I see the market changing course, abandoning X86 and going ARM. Two AMD's(meaning two companies without huge secured fab capacity at their disposal) can't sustain the x86 market.

 

[MacZ]

The joys of MBA leadership.
The gift that keeps on giving.

 

[londiste]

If Intel doesn't come out of this mess intact in a year or two, I see the market changing course, abandoning X86 and going ARM. Two AMD's(meaning two companies without huge secured fab capacity at their disposal) can't sustain the x86 market.

Why? x86 vs ARM does not have much to do with foundries. And I do bet that Intel as a last ditch effort - if that should really start to happen - will open x86 up, at least for licensing.

But honestly, we should probably pray that Intel gets their process nodes figured out. Or Samsung catches up or someone else comes in (SMIC?) which is very unlikely. There is TSMC, Intel is falling behind and Samsung has already fallen behind. Of course there is a big market even when on nodes that are not cutting edge and business won't stop for them but if we as enthusiasts want to have newer and cooler CPUs and GPUs there also needs to be both capacity and competition in the cutting edge of manufacturing space.

 

[Jism]

"It's been a difficult few weeks"

Ironic coming from a company who had total dominance for years.

 

[Assimilator]

@AleksandarK

Report: Intel Could Spin Out Foundry Business or Scrape Some Expansion Plans to Control Losses

scrap, verb: To discard; to get rid of.

 

[las]

If Intel doesn't come out of this mess intact in a year or two, I see the market changing course, abandoning X86 and going ARM. Two AMD's(meaning two companies without huge secured fab capacity at their disposal) can't sustain the x86 market.

Intel don't need to come out of this really, then they will rely 100% on TSMC, just like AMD does right now. Intel needs to keep their own fabs, open them up for business, which is the plan.

Generally, TSMC needs competition. Both Intel and Samsung chases them and are not that far behind actually. Tables should turn in the 2-4 years. Tons of US and EU companies wants chip production out of Asia. Thats the reason even TSMC is building US and EU fabs. Intel only builds in US an Europe for this exact reason.

Intel 18A is Intels most important node ever. Already sampling. Testing. Improving yields. This is equivalent of TSMC 3nm, maybe even 2nm (as in the optimized 3nm node)

Arrow Lake using 3nm TSMC is a stop gap solution before Intel will be up and running fully on 18A and beyond. AMD already loses "leadership" here, back to being behind on node, for the first time in many years. As both Zen 5 and Zen 5 3D will be using 4nm which is 5nm in reality.

TSMC is the reason Zen became a succes (eventually). Zen sucked when AMD used GloFo for chip production. Intel has been competing with node-disadvantage for years and still has competitive performance. How will it look when Intel gets node-advantage? We will see very soon. Arrow Lake releases in October. Zen 5 disappointed so far and 9000X3D was delayed till 2025.

Perfect timing for Intel.

Intel also delivers better all-round performance than AMD. AMD needs 3D cache to be competetive in gaming, sadly 3D chips are mediocre outside of gaming. Also dual CCD chips are mediocre for gaming, beat by the single CCD counterpart. This is the reason I own 7800X3D and not 7950X3D. Pointless. I only game on this rig. I do work on serverclusters, not at home.

Intels biggest problem has been power usage, which should be fixed by going TSMC 3nm and eventually 18A.

 

[john_]

Why? x86 vs ARM does not have much to do with foundries. And I do bet that Intel as a last ditch effort - if that should really start to happen - will open x86 up, at least for licensing.

It does. OEMs move the markets and we seen that even at times when AMD was offering the superior product, OEMs where staying fixed at Intel. Why? Because it could deliver the quantities OEMs wanted. As long as Intel is having those fabs working for them, X86 will remain the main architecture in the market, except in phones obviously. Of course Intel needs to catch up with TSMC in process nodes, because we are seeing for years now (AMD vs Intel) that architecture alone isn't enough when someone is 2-3 nodes ahead. The X86 market would have been completely different if AMD and Intel where at the same node. So, Intel needs a valid process near at TSMC's level and fabs. If they fail, OEMs will start going faster and faster to ARM solutions because X86 will have lost it's main advantage and that's Intel's capacity. When was Intel and X86 competitive to ARM in efficiency? When Intel had the advantage over TSMC and everyone else. When Intel had a clear process advantage X86 Atoms looked efficient enough compared to ARM SOCs. Intel was purring billions back then in promoting it's chips, hoping to win the tablet market. I guess they where sure they will retain that process advantage for many many years. When Intel lost the manufacturing advantage, it just abandoned any idea of fighting ARM in the tablet market. The same will happen if they sell their fabs and become just another customer of those fabs. OEMs will start looking at ARM SOCs as the best solution for them. A few steps back in performance that most people wouldn't realise, a few steps ahead in battery efficiency that most people would realise, probably cheaper platform costs.

But honestly, we should probably pray that Intel gets their process nodes figured out. Or Samsung catches up or someone else comes in (SMIC?) which is very unlikely. There is TSMC, Intel is falling behind and Samsung has already fallen behind. Of course there is a big market even when on nodes that are not cutting edge and business won't stop for them but if we as enthusiasts want to have newer and cooler CPUs and GPUs there also needs to be both capacity and competition in the cutting edge of manufacturing space.

I do want them to fix their problems. I do want Samsung to become competitive in process node with TSMC to see if AMD can become more competitive in the market, having more wafers at their disposale. I don't see SMIC doing anything in the next few years, because of US restrictions. Maybe in the next decade when they have figured out how to produce advanced equipment the way ASML does.

Intel don't need to come out of this really, then they will rely 100% on TSMC to deliver, just like AMD does right now.

AMD is relying on TSMC that's why it can't win OEMs. Because it needs to wait for TSMC to serv Apple, then Nvidia, then Qualcomm and then AMD. Now it's also intel in TSMC's catalog of customers and guess what. Intel enjoys higher revenues, meaning they can also pay higher prices than AMD for TSMC's wafers. So AMD goes even further down in the priority list.

 

[Dristun]

The best cost-cutting measure Intel can take right here and now is firing "advisors" from Goldman and Morgan Stanley, lmao. They'll spin and sublicense the entire corp away while giving money to these fellas in nice suits.

 

[Crackong]

This is fine.
Intel's stock not at $3 yet

 

[las]

/AMD is relying on TSMC that's why it can't win OEMs. Because it needs to wait for TSMC to serv Apple, then Nvidia, then Qualcomm and then AMD. Now it's also intel in TSMC's catalog of customers and guess what. Intel enjoys higher revenues, meaning they can also pay higher prices than AMD for TSMC's wafers. So AMD goes even further down in the priority list.

Apple will always be top-priority at TSMC. TSMC is where they are today, due to Apple money.

Nvidia don't even use TSMCs prime-nodes, even tho they can afford it. They will in 2025 tho. Blackwell - both AI and consumer - is 4nm again, just like 4000 series and Hopper. They don't need to rush to 3nm, they already dominate using a cheaper node.

The only companies to use 3nm TSMC is Apple and soon Intel.

I don't see AMD using 3nm before late 2025 / early 2026. They don't have the funds. Spent 5 billions on ZT Systems recently. They are chasing AI, trying to make a dime and moon like Nvidia stock.

RDNA5 on 3nm (or better) is the next big thing from AMD. Lets hope they can deliver 4090/5080 performane by then. They will not touch 5090 at all, even tho Nvidia gets a 1+ year headstart.

RNDA4 will be another joke. 4/5nm with mid-end focus.

 

[LittleBro]

Intel also delivers better all-round performance than AMD. AMD needs 3D cache to be competetive in gaming, sadly 3D chips are mediocre outside of gaming. Also dual CCD chips are mediocre for gaming, beat by the single CCD counterpart. This is the reason I own 7800X3D and not 7950X3D. Pointless. I only game on this rig. I do work on serverclusters, not at home.

Intels biggest problem has been power usage, which should be fixed by going TSMC 3nm and eventually 18A.

You've got it wrong. Brute forcing highest possible clocks is not a way to achieve all-round performance.
3D cache is great example of a way to increase performance by touching architecture. You get higher gaming performance than i9-14900K with 1/5th of wattage. That's the way.
It's not like AMD needs the 3D cache that saves them, otherwise Intel is best ... That can be said the other way around: Intel needs insane clocks, voltages and wattage to save them in gaming ...
It's about the innovation that AMD did and that brough them success. Hybrid architecture brought Intel success, but past Alder Lake those extreme wattages brought Intel only problems.

Dual CCD chips are not intended for gaming. AMD's gaming chips are (apart from 3D) x600X(T), x700X(T) and x800X(T).
I would not call anything in range of 7600X, 7700X, 9600X, 9700X, 7800X3D mediocre in gaming.
9600X sits just 5% below best that Intel has to offer, at a fraction of that i9's power and price.

Now with branch prediction fix it is even faster. Even 7600X might surpass i9-14900K's gaming performance while not degrading/burning out.
AMD chips have been crippled by 3 years now, because their architecture was not properly utilized by the OS. However, tests with Intel chips does not show the same result.

Arrow Lake is said to lower TDP (compared to 14th Gen) by a 20-25%. Intel needs invention that will help them 7800X3D performance at like 100 Watts or even lower.

 

[napata]

You've got it wrong. Brute forcing highest possible clocks is not a way to achieve all-round performance.
3D cache is great example of a way to increase performance by touching architecture. You get higher gaming performance than i9-14900K with 1/5th of wattage. That's the way.
It's not like AMD needs the 3D cache that saves them, otherwise Intel is best ... That can be said the other way around: Intel needs insane clocks, voltages and wattage to save them in gaming ...
It's about the innovation that AMD did and that brough them success. Hybrid architecture brought Intel success, but past Alder Lake those extreme wattages brought Intel only problems.

Dual CCD chips are not intended for gaming. AMD's gaming chips are (apart from 3D) x600X(T), x700X(T) and x800X(T).
I would not call anything in range of 7600X, 7700X, 9600X, 9700X, 7800X3D mediocre in gaming.
9600X sits just 5% below best that Intel has to offer, at a fraction of that i9's power and price.

View attachment 361313

Now with branch prediction fix it is even faster. Even 7600X might surpass i9-14900K's gaming performance while not degrading/burning out.
AMD chips have been crippled by 3 years now, because their architecture was not properly utilized by the OS. However, tests with Intel chips does not show the same result.

Arrow Lake is said to lower TDP (compared to 14th Gen) by a 20-25%. Intel needs invention that will help them 7800X3D performance at like 100 Watts or even lower.

Aggregate of average framerates at 1080p where multiple games are partially GPU bottlenecked is such a bad way to look at CPU performance. It's great for your argument but it always make top CPUs look worse than they really are (including the 7800X3D) and minimizes the differences.

From PCGH:

Quite the difference with your 5%, right?

 

[Daven]

Three major events contributed to the downfall of Intel:

Release of the Apple iPhone
Release of the Nvidia Tesla C870 GPU Computing Module
Release of AMD Zen architecture

Intel and computer enthusiasts alike never could understand the ramifications of the above three at the time. Some are still in denial.

Edit: More importantly, Intel cannot ‘win’ back it’s past market performance like it did after the release of the Core architecture. The fundamentals of the business have changed irrevocably and Intel doesn’t have a viable business strategy to run all those expensive fabs just for it’s own products.

 

[MacZ]

Intel and computer enthusiasts alike never could understand the ramifications of the above three at the time. Some are still in denial.

$100 billions in share buybacks since 2005 says otherwise.

 

[LittleBro]

Aggregate of average framerates at 1080p where multiple games are partially GPU bottlenecked is such a bad way to look at CPU performance. It's great for your argument but it always make top CPUs look worse than they really are (including the 7800X3D) and minimizes the differences.

Quite the difference with your 5%, right?

Intel has much better IMC, that's true. That is the biggest problem of Zen right now - shitty IMC.
You can see how X3D chips are better thanks to much larger cache, so they don't have to access DDR4/5 so often.

 

[ARF]

Long overdue. What goes around, comes around. You reap what you sow.

 

[usiname]

It does. OEMs move the markets and we seen that even at times when AMD was offering the superior product, OEMs where staying fixed at Intel. Why? Because it could deliver the quantities OEMs wanted.

Nope, the reason is the cancer (Intel) that infected the whole OEM industry blocking the competitors.

 

[las]

You've got it wrong. Brute forcing highest possible clocks is not a way to achieve all-round performance.
3D cache is great example of a way to increase performance by touching architecture. You get higher gaming performance than i9-14900K with 1/5th of wattage. That's the way.
It's not like AMD needs the 3D cache that saves them, otherwise Intel is best ... That can be said the other way around: Intel needs insane clocks, voltages and wattage to save them in gaming ...
It's about the innovation that AMD did and that brough them success. Hybrid architecture brought Intel success, but past Alder Lake those extreme wattages brought Intel only problems.

Dual CCD chips are not intended for gaming. AMD's gaming chips are (apart from 3D) x600X(T), x700X(T) and x800X(T).
I would not call anything in range of 7600X, 7700X, 9600X, 9700X, 7800X3D mediocre in gaming.
9600X sits just 5% below best that Intel has to offer, at a fraction of that i9's power and price.

View attachment 361313

Now with branch prediction fix it is even faster. Even 7600X might surpass i9-14900K's gaming performance while not degrading/burning out.
AMD chips have been crippled by 3 years now, because their architecture was not properly utilized by the OS. However, tests with Intel chips does not show the same result.

Arrow Lake is said to lower TDP (compared to 14th Gen) by a 20-25%. Intel needs invention that will help them 7800X3D performance at like 100 Watts or even lower.

Still, 7800X3D is top choice for gaming, yet performs worse than 7700X outside of gaming. I own a 7800X3D solely because of gaming perf, and will drop it for 9800X3D or Arrow Lake soon anyway.

With Intel you can get both good gaming perf and good application perf in the same chip. You don't have to choose one over the other. Downside is power usage, however, not in gaming. 14700K uses like 100 watts on average in gaming.

3nm TSMC will fix Intels power consumption issues, their performance have been fine even with AMD having node advantage. Lets see what Intel can do with node-advantage.

Arrow Lake and eventually 18A is going to be the turning point for Intel. AMD lost node advantage, and Zen 5 somewhat failed - However 9800X3D might take the gaming crown anyway, due to being unlocked for OC and running higher clocks in general. 3D cache is less fragile this time.

 

[Fatalfury]

Intel thought they could have one "14nm" making foundry for 10 years.. then 10nm for another 7 years then slowely go 5nm then by the time they reach 3nm it would be the year 2035.
too bad TSMC had other plans...

 

[Eternit]

Intel 18A is Intels most important node ever. Already sampling. Testing. Improving yields. This is equivalent of TSMC 3nm, maybe even 2nm (as in the optimized 3nm node)

Yes, but they must work on the next two nodes as well or the 18A would be like 14nm was. The question is can they afford to do that.

 

[Daven]

Nope, the reason is the cancer (Intel) that infected the whole OEM industry blocking the competitors.

^^This^^

We will never have a healthy innovative market that thrives off competition as long as one player that gets ahead becomes complacent and thinks they ‘deserve’ the market regardless of current and future performance.

Such a company truly becomes a cancer that takes enormous resources to cut out and suppress.

 

[las]

Yes, but they must work on the next two nodes as well or the 18A would be like 14nm was. The question is can they afford to do that.

Intel just got $8.5 Billion from Joe Biden

Big part of Intels financials are because they have been spending big on new fabs, which are soon up and running everywhere

Tables are turning very shortly, all big tech companies want chip production out of Asia, and Intel is the best bet in doing so, and they will soon open their foundries up for business

Intel can be making AMD CPUs and GPUs in a few years

 

[Daven]

Intel can be making AMD CPUs and GPUs in a few years

This will not happen. You really have to think about it. At best, AMD would be paying Intel to improve Intel products and put AMD out of business. At worst, Intel will sabotage AMD chip production and put AMD out of business. Intel’s past suggest they will do both and everything in between. Entertaining anything less malevolent is just wishful thinking based on brand loyalty.